Re: [AMBER] QUESTION ON MMPBSA

From: Jason Swails <jason.swails.gmail.com>
Date: Tue, 19 Apr 2011 20:18:51 -0700

On Tue, Apr 19, 2011 at 7:59 PM, vani panguluri <vanipanguluri.gmail.com>wrote:

> Hello,
>
> If we run the MPI version of sander and pmemd , we cannot use them in the
> serial version of MMPBSA.PY ???
>

MMPBSA.py calls only sander by default (not sander.MPI). We parallelized
MMPBSA.py by implementing MPI in the python script rather than calling
parallel programs. This gives better scaling for a large number of
snapshots (which is the only time it really makes sense to use the parallel
version, anyway).

You could hack MMPBSA.py to look for sander.MPI instead of sander and make
sure you modify the system call to use some kind of mpi command beforehand.
However, we used to use that approach and abandoned it. It's inconsistent,
and will not work on every system. Each MPI implementation each system
(especially supercomputers) differs, and it's unclear exactly how to use it
*properly* within a python script.

HTH,
Jason


> can anyone tell whether we can use them or not??
> Did I need to repeat all the steps again??
> I cannot upgrade the ambertools 1.5 because i am using LONI ACCOUNT....
>
> Regards,
> Vani
>
> On Sun, Apr 10, 2011 at 10:26 PM, Jason Swails <jason.swails.gmail.com
> >wrote:
>
> >
> >
> > On Sun, Apr 10, 2011 at 6:16 PM, vani panguluri <vanipanguluri.gmail.com
> >wrote:
> >
> >> Hello,
> >>
> >> using the serial version of MMPBSA.py i got the following error.
> >>
> >> ptraj found! Using
> >> /usr/local/packages/amber/11/intel-11.1-mvapich-1.1/exe/ptraj
> >> sander found! Using
> >> /usr/local/packages/amber/11/intel-11.1-mvapich-1.1/exe/sander
> >> ptraj found! Using
> >> /usr/local/packages/amber/11/intel-11.1-mvapich-1.1/exe/ptraj
> >> sander found! Using
> >> /usr/local/packages/amber/11/intel-11.1-mvapich-1.1/exe/sander
> >> ptraj found! Using
> >> /usr/local/packages/amber/11/intel-11.1-mvapich-1.1/exe/ptraj
> >> sander found! Using
> >> /usr/local/packages/amber/11/intel-11.1-mvapich-1.1/exe/sander
> >>
> >
> > This tells me that you're still using MPI commands to launch the serial
> > version of MMPBSA, so it's launching several MMPBSA.py serial jobs all at
> > the same time. This will cause massive problems, as each thread tries to
> > write to exactly the same files, and they will be fighting each other.
> >
> > Review running programs in both serial and parallel to make sure that
> > you're doing it properly.
> >
> >
> >> Calculations complete. Writing output file(s)...
> >>
> >> Error: Sander output is missing values!
> >> VDWAALS = ************* EEL = -5060.4993 EGB =
> >> -14223.5294NOTE: All files have been retained for
> >>
> >
> > This tells me that your files are incompatible (your prmtop files and
> > trajectory files). Make sure that you set up your calculation properly
> so
> > that the complex/receptor/ligand topology files match the generated
> > trajectory files (you can check this with visualization programs).
> >
> > Good luck,
> > Jason
> >
> > --
> > Jason M. Swails
> > Quantum Theory Project,
> > University of Florida
> > Ph.D. Candidate
> > 352-392-4032
> >
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>



-- 
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Apr 19 2011 - 20:30:04 PDT
Custom Search